Identification of flow regime in a bubble column reactor with a combination of optical probe data and machine learning technique

Abstract

In the present work, a data-driven model for identification of flow regime in a bubble column is developed by combining data from optical probe technique and machine learning. Optical probe data from previous work was combined with new data in the present work to expand the database for model development. A novel methodology for determination of two key parameters from the optical probe signal, bubble time and characteristic time of the signal, is presented. The significance of these two parameters is that they contain rich information on operating flow regime in the bubble column. A map of these two parameters for various operating conditions is created, showing points belonging to identical flow regime lie in a cluster. A machine learning methodology based on support vector analysis was developed to identify flow regime using map developed in this work. This approach was able to uniquely classify flow regimes for various experimental conditions on single map, which is the highlight of this work.